Improving laser image resolution for pitting corrosion measurement using Markov random field method

To characterize the pitting corrosion of metallic pipe, high-resolution laser scan is indispensable. In many cases, only low-resolution scan can be obtained due to the limitations of the scanning equipment or time constraint. Although interpolation method can be applied to enlarge the low-resolution image, the enlarged laser scan loses the details of surface topography, which are important to calculate the parameters of pitting corrosion. In this paper, a singe-frame super resolution method is proposed to infer a high-resolution laser scan from the low-resolution input. The relationship between the low-resolution input and high-resolution result is modeled with a Markov random field (MRF) with the aid of a training set built in advance. A belief propagation algorithm is implemented to infer the super-resolved result. The experiments demonstrate a good performance of the proposed method in comparison with the traditional interpolation methods.